The Impact of Private Schools, School Chains and PPPs in Developing Countries Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Lee Crawfurd , Susannah Hares, and Rory Todd The private school sector has expanded with almost no public intervention to educate half of primary school children in many urban centers in Africa and Asia. Simple comparisons of test scores would suggest that these private schools may provide better quality than public schools, but how much of this difference is due to selection effects is unclear. Much donor and policymaker attention has proceeded on the basis that private schools do perform better, and focused on models of public subsidy to expand access, and investment in networks or chains to encourage expansion. We review the evidence of the effects of private schools on learning, and how that effect translates to public-private partnerships (PPPs). We also study the effects of private school chains. We conduct a systematic review for eligible studies, with transpar- ent search criteria. The search resulted in over 100 studies on low-cost private schools and PPPs, with a large majority being on low-cost private schools. Our meta-analysis shows moderately strong effects from private schooling, although the limited number of experi- mental studies find much smaller effects than quasi-experimental studies. This advantage, though, is not nearly enough to help most children reach important learning goals. Turning to policy goals, we find that the private school advantage has not translated to public pri- vate partnerships, which have shown limited value in improving quality. They can however represent a lower-cost means of increasing access to school. We also find that private school chains perform little better than individual private schools and have little scope for achieving meaningful scale. JEL Codes: O1, I2 Keywords: chains, developing countries, private schools, PPPs. Introduction Nearly 100 million children attend private primary schools, the majority of whom live in low- and middle-income countries. This is equivalent to 19 percent of primary school students in low- and middle-income countries. This ranges from 14 percent The World Bank Research Observer © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com https://doi.org/10.1093/wbro/lkad005 39:97–123 in sub-Saharan Africa and 19 percent in Latin America, to 38 percent in South Asia (up from under 20 percent in 2000).1 In many major cities in Africa, Asia, and Latin America, over half of children in primary school attend a private school. Enrolment in private schools is typically even higher at preschool (41 percent of enrolment in low- and middle-income countries) and secondary school (28 percent of enrolment). Since large shares of children now attend private schools, it is important to understand Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 their quality. This also makes more salient the question of whether governments can and should use the private sector to deliver publicly-financed education. Public funding for places in existing private schools may in some cases be a viable strategy to rapidly expand ac- cess at lower cost than through the public sector (Barrera-Osorio et al. 2020, 2022). This strategy may be particularly attractive for secondary education or pre-primary education where enrolment is still low. Supporters of PPPs also suggest they can im- prove quality through better management or increased competition, with the public sector incentivized to react in education systems funded through per-capita payments (Friedman and Friedman 1962). Others highlight risks, including the challenge for governments to supervise a mixed sector and the potential for PPPs to further seg- ment education systems leading to increased inequity (Patrinos et al. 2009). In this paper we review three strands of recent literature on private schools. First, we review the evidence on the effectiveness of private schools in developing coun- tries at improving student outcomes, updating a prior review (Ashley et al. 2014). Second, we focus on those private schools that are part of a network or chain. Third, we consider the effectiveness of private schools that receive public funding as part of a public-private partnership (PPP) arrangement, updating another earlier review (Aslam et al. 2017). We find that the average private school advantage over public schools is posi- tive and moderately large compared with other interventions (though noting that most other interventions are much less intensive). This advantage, though, is not enough to help most children reach important learning goals. The gap between pri- vate and public schools drops sharply after controlling for important unobserved fac- tors through randomization. We find just five studies that compare the performance of private school chains or networks with other private or government schools in developing countries. We com- plement this literature review with original analysis of five datasets on school chains. Overall the evidence suggests that schools in chains or networks achieve similar out- comes to independent private schools. Schools in chains also have similar variation in quality to independent private schools. Finally, we see little scope for any individ- ual chain to reach sizable scale, ultimately therefore remaining peripheral to public education systems. Turning to public-private partnerships, we find just five additional rigorous studies since the 2017 review. The new evidence is consistent with the conclusions of the ear- lier review—it is inconclusive and context specific but it is possible to conclude that 98 The World Bank Research Observer, vol. 39, no. 1 (2024) the private school advantage is not yet translating into effective PPPs. Undertaking PPPs in weak or fragile states seems particularly challenging (Romero and Sandefur 2019, Romero et al. 2020). It is exactly those contexts where government capability to directly run schools effectively is weakest, that government capability to success- fully and sustainably manage contracts with private partners is also weak. In the next section, we discuss results on private school effectiveness, followed by those on school Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 chains, then those on public-private partnerships, before concluding. Private Schools Study Selection We conduct a systematic review and meta-analysis of empirical studies on the ef- fectiveness of private schools in developing countries published between 2008 and 2020. We begin with the 20 studies reviewed by Ashley et al. (2014) between 2008 and 2014, adding to these through a comprehensive search for studies published be- tween 2014 and 2020. We searched Google Scholar for the following search terms combined with the name of each low- and middle-income country: Private schools; privatization; school competition; low-fee private schools. We also searched the Africa Education Research Database and the NBER working paper series. The search was conducted between September and October 2020. Our initial search yielded 328 studies. We then limit these studies to those that include measures of student learning outcome, and have an experimental or quasi- experimental research design. To be eligible for meta-analysis, studies needed to pro- vide an effect size or the data required to calculate an effect size expressed in standard deviations (SD), as well as a standard error for this effect. Where studies give results for different outcomes, for different regions, or for different school types (registered or unregistered private schools), and with no aggregated effect size, we extract all re- sults. Where studies provide effect sizes for multiple time periods or follow-ups, we use the effect size for the longest duration. Where studies provide estimates using multiple research designs, we select the estimate with the strongest claim to causal inference. For the main result, we include only studies with the strongest research designs to identify causal effects. In our assessment this includes Randomized Control Trial (RCT) studies which randomly assign vouchers to study at private schools, and “value-added” model studies which use longitudinal student data to control for prior achievement.2 We use random effects to separate variation in true effect sizes from sampling error. We run a sub-group analysis to estimate differences between effect sizes from RCTs and value-added studies. We also compare differences between stud- ies from the earlier review (Ashley et al. 2014), and those from this updated review (fig. A1.1). It was possible to extract 22 results in total, from 12 different studies. These estimates came from three studies from Ashley et al. (2014), and a further nine Crawfurd et al. 99 Figure 1. Meta-Analysis of Private School Effects on Learning Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: Where we have extracted multiple results from a study, the effect given is the fixed effect for that study in a model with fixed effects for each study. studies from our own review. Of the twelve studies, four were RCTs, and eight were value-added studies. Our sample is skewed towards India, with seven out of twelve studies conducted there. There is one study from Pakistan, one from Uganda, one from Tanzania, one from Ethiopia, and one from Peru. Effectiveness of Private Schools Results are shown in the forest plot in fig. 1, disaggregated by method. The mean of true effect sizes was estimated as 0.26 SD, with confidence intervals at 0.14 SD and 0.37 SD. This is a moderately large effect size compared with other interventions studied in low- and middle-income countries (Evans and Yuan 2022). The I-squared, which gives the percentage of the variation in effect sizes that can be explained by 100 The World Bank Research Observer, vol. 39, no. 1 (2024) true variation as opposed to sampling error, was 83.4 percent, representing strong heterogeneity of effect sizes. This may indicate that the effect of attending private schools varies significantly, depending on context and implementation quality, or al- ternatively that the different methods used for estimation have an important influ- ence on estimated effect sizes. Effects from RCT studies were smaller than value-added studies, with a test for dif- Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 ferences between the two yielding a p-value of 0.08, significant at the 10 percent but not 5 percent level. This may suggest that value-added studies overestimate the effect of private schooling on learning due to their additional assumptions not being up- held. However, as we only have four RCT studies—all of which are in India, and two of which study the same experiment—this interpretation is tentative. We also note that comparisons of standard deviation effect sizes across studies should be treated with caution (Singh 2015a). A regression-based Egger test for small-study effects had a p-value of 0.34, indicating a lack of evidence for small-study effects and therefore a lack of evidence for publication bias. How big are all of these (relative) effect sizes in absolute terms? Few papers report outcomes on a more tangible scale than “standard deviations of learning.” In Delhi, where a negative effect on learning was found, a 0.09 standard deviation reduction in test scores due to winning a voucher was equivalent to being able to read 4 fewer words per minute—compared to an average of 47 words (Crawfurd et al. 2021a). One observational study (Alcott and Rose 2016) reports that children who attended private school were 6–8 percentage points more likely to have learnt “the basics”— defined as being able to read a paragraph and do basic multiplication—by the end of primary school. This still leaves a large share of students, particularly poorer ones, unable to complete the basics. Overall, achieving the large improvements in learn- ing necessary to meet basic learning goals is likely to need the stacking of multiple successful interventions (Eble et al. 2020, Fazzio 2021). Why is there such strong demand for private schools among parents in lower- and middle-income countries, given only modestly better test scores? A few studies in our review address this question. Karopady (2014) collects qualitative evidence to exam- ine motivations for parents sending their children to private school in rural India. Parents mentioned the use of English as a key incentive, as well as the presence of smart uniforms, the higher social class of other pupils, and the greater amount of homework and school hours offered. Crawfurd et al. (2021) test explicitly for the pres- ence of such non-learning benefits to attending private schools, but find no effects. They do however find explicit evidence for parents overestimating the returns to pri- vate school, before enrolment. This view of information asymmetry is supported by Eigbiremolen et al. (2020), who argue that the lack of regulation of private schools in Peru causes parents to unwittingly pay far over the odds for low-quality private schools. Crawfurd et al. 101 Table 1. Violence Against Children in Private Schools Private– government Government Private gap Percentage of 15-year-olds reporting sexual harassment at school by a teacher or other staff member in last four weeks Cambodia 4.2 4.1 −0.0 Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Ecuador 0.9 1.1 0.2 Guatemala 2.9 1.7 −1.1 Honduras 2.4 1.6 −0.9 Paraguay 1.3 0.4 −0.9 Senegal 13.6 13.6 −0.0 Zambia 15.4 4.5 −10.9 Percentage of 13–24 year olds who have ever been punched, kicked, whipped, or beat with an object by a teacher Haiti 18.8 20.8 2 Kenya 68.6 47.0 −21.6 Tanzania 64.8 71.3 6.5 Note: Data on sexual harassment is from the 2017 PISA for Development survey, and data on physical violence is from the Violence Against Children Surveys—2012 in Haiti, 2010 in Kenya, and 2009 in Tanzania. Mechanisms and Other Outcomes Test scores are not the only outcome that matters. Safety is a high priority for parents and students. What limited data exists suggests that overall, children are exposed to similar levels of physical and sexual violence in public and private schools, with some variation across countries (table 1). If private schools do perform better on average, how much of this better perfor- mance is due to greater financial resources, and how much is due to better school management? Across low- and middle-income countries, private schools perform slightly better than public schools on average in terms of their management prac- tices. This is shown by Crawfurd 2017 in Uganda, Lipcan et al. (2018) in Nigeria, and Scur et al. in India. In addition, we use data from the PISA survey to calculate a management index, following Leaver et al. (2019). According to this data, private schools score 0.1 SD better on management practices, though this difference is not statistically significant (supplementary online appendix fig. A2). Implications for Equity A common concern about private schooling in the developing world is that it under- mines equality in access and attainment for children. In other words, a concern that 102 The World Bank Research Observer, vol. 39, no. 1 (2024) private schools improve learning outcomes for richer children more than poorer chil- dren. One argument is that private schools exist more in urban areas where wealth is concentrated and more parents have the ability to pay, leaving poorer rural areas underserved. Even where private schools are prevalent in rural areas, they may not be affordable to poor families. Many studies have attempted to tackle this question by focusing on the accessibility and affordability of private schools for poor children Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 (Ashley et al. 2014; Nguyen and Raju 2015; Harma 2016; Srivastava and Noronha 2016; Baum et al. 2018; Patel and Sandefur 2020). A related but distinct concern is unrelated to learning and school quality alto- gether, but instead whether private schools increase the degree of socio-economic segregation between schools. Aside from any difference in quality, private schools could lead to greater overall socioeconomic segregation, with overall negative wel- fare consequences. Whilst some have measured the relative amount of segregation within the private sector and within the public sector (Siddiqui 2017), we have not identified studies that address the question of the extent of the private school market on overall school segregation. This matters, as it is perfectly possible that segregation can be just as severe in an entirely public system as an entirely private one, due for example to strong residential sorting (e.g., Monarrez 2022) if rich children attend different public schools to poor children. More research would help answer this ques- tion and in the meantime we do not think it is possible to conclusively say whether private schools increase inequity or not. A third related concern is around whether a growing private sector has negative spillover effects on students remaining in the public sector. Three studies address this question directly, showing no negative impacts on children remaining in the public sector from a growth in private schools (Muralidharan and Sundararaman 2015; Joshi 2020; Bagde et al. 2022). In summary, our meta-analysis shows that on average private schools perform 0.26 SD better than public schools on test scores. These improvements in test scores are mirrored by similar improvements in management practice. However the practi- cal implications of these effect sizes are relatively small, and private schools do not perform significantly better at reducing violence in schools. Finally regarding strati- fication and segregation, there is little evidence on whether the emergence of private schools increases overall segregation between all schools within a school market. Private School Chains In this section we review the evidence on the potential for growth, quality, and consis- tency of school chains. We define chains broadly, to include both for-profit corporate groups, and non-profit charitable networks of schools. We argue that private school chains play an oversized role in public discourse around private education compared Crawfurd et al. 103 to their actual role. Evidence on their effectiveness is limited, and their potential for scale seems small. Philanthropists and official Development Finance Institutions have often in- vested in private school chains. Of the Development Financial Institutions, the World Bank’s International Finance Corporation (IFC) is by far the largest in- vestor in education, with over $1 billion invested. The majority of IFC invest- Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 ments on education—$811 million—have been in higher education or adult train- ing. Within the school sector, around half ($181 million) has gone on school chains (supplementary online appendix fig. A3), with this number rising in re- cent years (Smith and Baker 2017). Most of these investments are, however, tar- geted at elite institutions that are out of the reach of most parents in low-income countries (Mundy and Menashy 2013). The United Kingdom’s development fi- nance institution, the CDC, made five education investments between 2012 and 2017, of which two were in Bridge International Academies, one in the low-cost brand from elite private school chain GEMS, and two in the private university UNICAF.3 Philanthropists and investors have argued that private school chains can be both attractive private investments, and valuable supplements to public education (Assomull et al. 2015; Featherston and Ferreira 2017; Abdo et al. 2019; Istance and Paniagua 2019). Critics on the other hand argue that the “commercialization” of education is highly dangerous. For example the UN Special Rapporteur on the right to education wrote that “commercial school chains thus threaten not only to impair the realization of the right to education, but also to damage democracy, social cohesion and stability in developing countries” (Barry 2019). Despite this attention from funders and investors and critics, we conclude that school chains have low relevance in education systems. Their overall market shares are low, with a low likelihood of substantial growth (with a few exceptions, such as BRAC in Bangladesh and The Citizens Foundation (TCF) in Pakistan, which together educate more than 200,000 children). Whilst detailed data on overall funding provided to school chains is not available, it’s likely that the funding they receive is disproportionately large given that the vast majority of children who attend private schools do not attend one run by a chain. Study Selection We begin by following the same search protocol as outlined in the previous section, with the search keywords “school chain; school network; school franchise;” in the pe- riod 2000–2020. This search returns just two quasi-experimental studies that con- sider the role of private school chains, and so we also consider three observational studies. We then complement this review with new analysis of data on school chains 104 The World Bank Research Observer, vol. 39, no. 1 (2024) from the United Kingdom, United States, Bangladesh, India, Pakistan, Nigeria, and Uganda. Effect of Chains on Average Learning Outcomes We know little about whether chains outperform individual private schools in test Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 scores, since few studies have made this comparison directly. For those that have, there is no indication of consistently better results from chain schools. We find just five studies that compare the performance of private schools in devel- oping countries that are members of chains or networks with other private or gov- ernment schools. Four of these are international chains, with just one paper cover- ing national chains in Chile. A United Kingdom-based NGO which runs 28 secondary schools in Uganda (“PEAS”) performs slightly better than other (local) public-private partnership schools in terms of test score value-added (Crawfurd 2017), and simi- larly to other low-cost private schools (Economic Policy and Research Centre 2016). Rising Academies, a network of 39 schools educating 7,800 pupils performs signif- icantly better than other private schools in Sierra Leone in both reading and math (Johnson and Hsieh 2019). In Lagos, Nigeria, Bridge schools perform better than sim- ilarly priced local private schools in reading, but no better in mathematics (Crawfurd and Lipcan 2018). Franchises in Chile get better results than independent private schools (Elacqua et al. 2011). As described in more detail in the conclusion, out- sourced schools run by chains in Liberia (Romero et al. 2020) perform slightly better than government schools but at high cost and with some harmful effects. It is difficult to compare the cost of educating a child in a school run by an international chain or an NGO with other low-cost private schools, government schools, or PPP schools, since NGOs generally do not factor in international headquarters and management costs and their “back office” into these calculations. There are also six studies from the United States and United Kingdom. US charters in chains do only very slightly better (Woodworth et al. 2017), and in the United King- dom, academy chains do no better on aggregate than government schools (Andrews 2016; Hutchings and Francis 2017). Other studies have shown that some individ- ual charter management organizations in the United States can improve student out- comes relative to a counterfactual (particularly those that practice the “No excuses” philosophy—Angrist et al. 2013), but these are not representative of all charter man- agement organizations (Angrist et al. 2010; Abdulkadiro˘ glu et al. 2016). Effect of Chains on Variation in School Quality Whilst performance on average test scores is inconclusive, an attractive feature of private school chains could in theory be relative consistency in performance. The Crawfurd et al. 105 Table 2. Consistency in Quality of Private School Chains Standard deviation of performance Chains Individual schools Nigeria (Bridge) 0.51 0.89 Uganda (PEAS) 1.19 0.89 Pakistan (Various) 0.80 1.02 Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 United Kingdom (Various) 0.80 0.92 Note: This table shows the standard deviation of school-average mathematics test scores. In Uganda and the United Kingdom, school performance is estimated using longitudinal student test scores, adjusting for prior performance. Due to data limitations, for Nigeria, school quality is estimated controlling only for parental income, and for Pakistan average test scores are used with no controls. Data for Nigeria is from Crawfurd and Lipcan (2018), and compares the variation in average performance of Bridge schools with the variation of performance in a matched set of pri- vate schools. For Uganda data comes from Crawfurd (2017), and compares variation of PEAS schools with other individual PPP schools. For Pakistan, data is from Crawfurd and Alam (2022) and compares variation within PPP school chains with variation of individual PPP schools. Finally, data for UK academies is from www.compare-school- performance.service.gov.uk and compares variation in performance of multi-academy trusts with variation across individual academy schools. support provided by back-office functions such as curriculum development, teacher training, and finance, could provide more consistency in quality. However, empirically our analysis finds that private schools in chains or networks are no more consistent than independent private schools in the same markets. Table 2 compares the range of performance of schools in various chains with other nearby private schools. On first inspection, the chains do seem to have lower vari- ance in performance than other schools. Bridge schools in Nigeria have less variance between them than a matched sample of nearby private schools. PEAS schools in Uganda have more variance than other publicly subsidized private schools. In the United Kingdom and Punjab, Pakistan, we have data on multiple chains. The average within-chain variation in performance is lower in both the United Kingdom and Pak- istan than other similar privately-managed public schools that are managed by sole operators. However on closer inspection much of this difference can be attributed to the relatively small size of chains. As chains grow past 30 schools in the United King- dom or 100 schools in Punjab, they converge on the same variation between schools as exists amongst sole operators (fig. 2). Potential for Scale of Chains In this section we analyze the existing size of private school chains in developing and developed markets. The regularity in the distribution of chain sizes gives us an in- dication of the current scale of school chains, as well as what could reasonably be expected to develop in the future. That we see limits to the scale of any one chain 106 The World Bank Research Observer, vol. 39, no. 1 (2024) Figure 2. Schools in Chains Vary in Quality as Much as Independent Schools Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: The left panel shows variation in performance within school chains in Punjab, Pakistan, by size. As chains grow in size, their variability approaches the level of overall variation in the market. Similarly in the right panel, variation within multi-academy trusts in the United Kingdom converges on the overall level of variation in the market as networks grow in size. Data for Pakistan chains is from Crawfurd and Alam (2022). Data for UK academies is from www.compare-school-performance.service.gov.uk. even in mature markets (such as private preschools in the United Kingdom) suggests that the underlying economics are not conducive to large private organizations run- ning a large proportion of schools in any single market—perhaps due to a lack of economies of scale in teaching and learning (Pritchett 2014), or due to the presence of information asymmetries in the measurement of output. First, we consider the current size of private school chains in the context of average public sector operating units (typically a district). Our data on the size of chains in de- veloping countries comes primarily from the Global Schools Forum (GSF)—a member organization for private school chains and other non-state organizations in develop- ing countries.4 According to 2021 data, it has 50 school chain members which run a total of 17,500 schools. The majority of these schools (9,000) are run by one mem- ber, BRAC.5 Private school chains are dwarfed in size by government school districts. The median chain (in the Global Schools Forum) runs just 10 schools. Only six chains can compete in size with the average school district in South Africa, Punjab (Pakistan), Kenya, and Uganda (supplementary online appendix fig. A4). And there are hundreds of government school districts in each country. Second, we consider the scope for growth of private school chains by looking at historical growth of international school chains. The median chain is 11 years old and has grown by 1.2 schools each year. There is a wide distribution, with a few very large chains that have grown quickly (BRAC has grown at an average of 244 schools per year; TCF at 57 schools) and other chains that are several years old but never grew beyond their initial one or two schools. By comparison, in India alone the overall number of private schools grew by over 15,000 per year between 2011 and 2016 (Kingdon 2020). Crawfurd et al. 107 Figure 3. School Chains Follow a “Power Law” Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Sources: Data for Global Schools Forum is from globalschoolsforum.org. Data for UK Preschools is from daynurs- eries.co.uk. Third, we consider the scope for chain growth by looking at the size of chains in more mature markets. The market for preschool in the United Kingdom is mostly pri- vate. Here, 60 percent of preschools are part of a chain. But the median chain size is only around five preschools. For UK primary and secondary academy schools and US charter schools, less than one in four are part of a chain, and the median chain has four schools. So even in healthy, developed markets with good rule of law, and good access to credit, land, and skilled labor, very few chains grow very large. The situation is similar in developing countries. A study of the private preschool market in four African cities found that over 80 percent of private schools in Accra, Lagos, Johannesburg, and Nairobi are independent (Bidwell and Watine 2014). Of the 20 percent that are in a chain, the median chain has only three or four schools (supple- mentary online appendix fig. A5). Fourth, we go beyond average size to look at the full distribution of chain sizes, to see to what extent that distribution is “normal.” Economists have extensively docu- mented the existence of “power laws” in the distribution of economic activity (Gabaix 2009). Across time and space, there is a strong empirical regularity in the size distri- bution of firms, with just a few very large units and a long tail of smaller ones. Graph- ically, this can be seen by plotting the logarithm of a unit’s rank against the logarithm of its size. We see the same clear pattern for UK preschools and for developing coun- try private school chains (fig. 3). The precise mechanism driving this distribution is unclear, but it is consistent with the growth of any individual unit being as good as random. All of this gives us little reason to believe that there will ever be more than a few very big global chains. Overall the data suggest that school chains are not, and will not become, a sub- stantial component of public education systems and will be peripheral to solving the twin challenges of getting kids into school and getting them learning. There are 108 The World Bank Research Observer, vol. 39, no. 1 (2024) exceptions, notably TCF and BRAC, two local organizations in Pakistan and Bangladesh respectively, which are running large networks of schools either in par- allel to government systems or as part of the government system. Their contribution to, and stake in, education in those countries is substantial and—as they become “too big to fail”—attention needs to be paid by the government to their performance and compliance with regulation. Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Beyond BRAC and TCF, most chains show limited potential to scale. In contrast, the independent low-cost private school market has already scaled, with tens of millions of children already enrolled across the developing world. Public-Private Partnerships Public-Private Partnerships (PPPs)—government funding for places in privately-run schools—are common in every region of the world. In this section we review the ev- idence of the impact of PPPs and discuss whether they have the potential to improve learning and equity at scale. We focus on PPPs in which governments provide funding for children to attend privately operated schools, rather than provisions of ancillary services in schools. To understand whether there is a strong evidence basis for undertaking a PPP in education, it is important to answer two questions. First, whether private schools can deliver better outcomes, as we discussed above. Our assessment is that the average effect of private schools is moderately large in comparison with most education inter- ventions, though much smaller after controlling for family background and impor- tant unobserved factors. Second, and the focus of this section, can better outcomes from private schools be replicated in a PPP arrangement at a cost that is affordable to the government? If governments cannot successfully and sustainably partner with the private sector—which includes running the procurement, monitoring and evalu- ation, and overall governance of a public-private partnership—more effectively than it manages its own schools, PPPs are not a viable policy option. While the promise of PPPs is attractive, the reality of implementation in weakly governed states is challeng- ing and governments may struggle to finance, monitor, and oversee private operators effectively, and may be unable to sanction or terminate contracts on the basis of poor performance. Study Selection To answer these questions, we build on three previous reviews. Patrinos et al. (2009) and LaRocque (2008) summarized the evidence on PPPs prior to 2009. Aslam et al. (2017) then summarized evidence on PPPs from 2009 to 2016. We begin by follow- ing the same search protocol as outlined in the previous section, with the following search keywords: public private partnerships; PPP; voucher schools; contract schools; Crawfurd et al. 109 concession schools; subsidy schools. We focus on the period since the last major re- view in 2017. We find just five rigorous studies published since 2017, and so we do not conduct a formal meta-analysis due to the small overall number of studies. Effects on Learning Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 The earlier two literature reviews found relatively few empirical studies examining the impact of Public-Private Partnerships (PPPs) on learning. They found some evi- dence that contract schools have a positive impact on student test scores, using studies from the United States, Venezuela, and Colombia, although the drivers of that suc- cess were unclear. The number of studies had grown by 2017, but remained context- specific and design-specific. Aslam et al. (2017) identified 22 new rigorous studies on PPPs; three on contract schools and nine each on vouchers and subsidies. The limited available evidence on contract schools suggests that they may be able to reach disad- vantaged students and their better management practices could lead to better out- comes. The evidence from the nine studies on subsidies suffers from methodological limitations but somewhat supports the claim that these programs can reach poorer children and have the potential to improve learning outcomes. Six of the nine studies on voucher programs are on Chile, from which the evidence is mixed and highlights the risk of increased inequity. Evidence from other contexts (India and Pakistan) is also inconclusive. Aslam et al. (2017) summarize the body of evidence on voucher provision as mixed, inconclusive and insufficient. Of the five studies published since 2017, three are on contract schools, one on vouchers and one an evaluation of Punjab, Pakistan’s PPPs, which includes contract, voucher and subsidy arrangements (supplementary online appendix table A1). What is clear from both the recent and older literature is that design matters. Operators within Public-Private Partnerships are not homogenous and PPPs them- selves are not homogenous. Table 3 shows the range of design features in PPPs, from contract schools that tend to use government teachers and government buildings through to voucher and subsidy programs that are much more “private.” Programs that involve paying existing private schools to educate children are generally cheaper for the government, since teacher salaries are typically lower. Beyond this, the PPPs in Sindh (Pakistan) and Uganda have successfully increased enrolment without hav- ing a negative effect on other outcomes, and successfully incentivized private en- trepreneurs to start schools in underserved areas. Case Studies: Liberia, Pakistan, Chile, Uganda In this section we discuss the design features of PPPs in Liberia and Pakistan and consider their long-term viability as part of the public education system. In particu- lar, we discuss whether PPPs in weaker states are too risky, given the limited capacity 110 The World Bank Research Observer, vol. 39, no. 1 (2024) Crawfurd et al. Table 3. Design Features and Impact of Different PPPs Colombla UK Uganda Liberia PSL Concessions Academies US Charters Punjab PSSP Sindh PPRS Chile SEP India RTE Secondary Year started 2016 2000 2001 1991 2016 2007 2008 2012 2007 Number of schools 194 25 7500 7 4300 2314 7500 c. 91,000 800 Type of PPP Contract Contract Contract Contract Contract Subsidy Voucher Subsidy Subsidy management management management management management Design No fees for top ups Yes Yes Yes Yes Yes - - Yes - Non-profit Yes Yes Mixed - - - Non-selective Yes Yes Yes Yes Yes Yes Yes _ _ Teachers on Govt contracts Yes - Mixed - - - - - - Unionized teachers Yes _ Yes _ _ _ _ _ _ Accountable for outcomes Yes Yes Yes Yes Yes Yes Yes National curriculum Yes Yes - Mixed Yes Yes Yes Yes Yes Government buildings Yes Yes Yes Mixed Yes - - - - Impact Cost compared with govt schools Higher Equivalent Equivalent Equivalent Lower Lower Lower Lower Lower Enrolment Lower Higher No effect No effect Higher Higher No effect No effect Higher Learning outcomes Small effect Small effect No effect No effect Small Higher Higher No effect Small effect negative effect Equity Negative Unclear No effect No effect Unclear Positive Positive Unclear Positive More More Public Private Note: This table summarizes the design features of different public-private partnership schemes. Sources include; Barrera-Osorio et al. 2007; Barrera-Osorio et al. 2016; Andrews and Perera 2017; Andrews and Perera 2017; Barrera-Osorio 2017; Cheng et al. 2017; Cheng et al. 2017; Crawfurd 2017; ; Damera 2017; Neilson et al. 2019; Romero and Sandefur 2019; Romero et al. 2020; Crawfurd et al. 2021. 111 Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 of weaker governments to hold private operators accountable for their performance; and whether different types of PPP are a financially sustainable policy option for gov- ernments. Liberia Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Two papers document a randomized control trial of Liberia’s PPP experiment, pro- viding a sobering account of the challenges of outsourcing education through con- tract schools in a fragile state (Romero and Sandefur 2019; Romero et al. 2020). In 2016, the Liberian government contracted eight private operators—a mix of local and international and non-profit and for-profit organizations—to manage 93 pub- lic primary schools under a PPP originally named Partnership Schools for Liberia (PSL) and now named the Liberian Education Advancement Program (LEAP). LEAP schools remained fee-free and non-selective, and LEAP teachers remained on govern- ment teacher contracts. Some learning gains were made—0.18 SD after one year and 0.16 SD over three years (equivalent to four words per minute additional reading flu- ency for the cohort that started in first grade) across the eight operators. Costs were high. In typical Liberian public primary schools, approximately $50 is spent on each student each year. Under LEAP, costs after one year ranged from an additional $57 per pupil to more than $1,000 per pupil. After three years this reduced to an average of around three times the government’s cost per child. Some negative effects were detected. LEAP reduced enrollment, driven by mass ex- pulsion by Bridge International Academies—a for-profit American company—who pushed excess students out of their schools and into other public schools. These stu- dents demonstrated a lower transition rate from primary school to secondary school. LEAP reduced corporal punishment from 51 percent to 46.4 percent but failed to re- duce sexual abuse. Of the eight operators, some produced wholly positive results while others did not. This study highlights the risks that policymakers in weak states should consider when minded to embark on a contract school PPP. Many of the challenges that emerged were as a result of gaps in the contract between government and private operators, which in itself was a reflection of weak governance and limited capac- ity of the state to negotiate with external organizations. The multiple scandals re- lating to charter schools shows that effectively commissioning, monitoring, and— where necessary—terminating private operators is challenging even for much better resourced governments and it may be impossible to fully capture all the mechanisms required. Beyond this fundamental challenge of contracting, the Romero et al. (2020) research shows risks relating to: a) financial sustainability: contract school programs may be simply unaffordable for governments. Since contract schools generally involve the private operator taking 112 The World Bank Research Observer, vol. 39, no. 1 (2024) on public school teachers, payroll costs—well over half of the total cost of educat- ing a child in almost every country—are unlikely to reduce. While self-reported costs of private operators in Liberia did reduce over time, schools run by the lowest- cost operators had approximately double the amount of money spent on them as typical government schools (supplementary online appendix fig. A6). b) promoting equity: LEAP concentrated its additional resources on a narrow set of Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 schools that were already advantaged. Schools were chosen to meet criteria named by operators, including number of teachers and distance from a road. Additionally, the allocation of more and better teachers to LEAP schools benefited those schools at the expense of other public schools; and c) heterogenous private operators: the differential results achieved by the eight differ- ent operators is an important policy consideration. Selecting and commissioning private operators for school management under similar contracts and in similar settings does not generate consistent results. The identity of the operator matters a lot and ex-ante it is challenging to predict which will produce uniformly posi- tive results and which will present stark tradeoffs between learning gains, access to education, child safety, and financial sustainability. Punjab, Pakistan Punjab, Pakistan has been a site of experimentation with PPPs for decades. The Punjab Education Foundation (PEF) was originally established in 1991 and sub- sequently revitalized in 2004 as a semi-autonomous organization attached to the Punjab School Education Department, tasked with promoting quality education through non-commercial public private partnerships. PEF deploys various PPP mod- els (see supplementary online appendix table A2) including voucher programs, a new schools program, and a contract management program for public schools (which has since been moved out of PEF into the direct control of the School Education Depart- ment). This section describes recent evidence on Punjab’s PPPs. Crawfurd (2018) studies one of the largest single education outsourcing exer- cises ever to take place. In Punjab, Pakistan, 4,276 underperforming public primary schools (around 10 percent of the total number of public schools) were contracted out to private operators in a single school year through the Punjab School Support Program (PSSP). Schools were defined as underperforming if they met a set of crite- ria primarily related to low enrolment or low standardized exam scores. Schools contracted to organizations who ran at least ten schools each numbered 2,600; the remaining 1,700 schools were contracted to individuals who ran a single school. The organizations and individuals were paid substantially less than the gov- ernment cost-per-child to run these schools. Unlike LEAP, Liberia, operators were able to hire their own teachers at wages set by the operators and existing public school teachers were offered the option of transferring to other public schools. Crawfurd Crawfurd et al. 113 found a large increase in enrolment, concentrated in Katchi (Kindergarten), although was not able to say with confidence whether this was driven by children who were al- ready enrolled in another school. In other words, it is not possible to conclude that PSSP increased overall enrollment. Overall there was a decline in exam scores in con- verting schools, though this may have been partly driven by the entry of new, lower- performing students. So it is not possible to determine whether this decline is due to a Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 negative treatment effect or the new composition of lower attaining students. Craw- furd notes that should the program expand further, effort should be made to identify and track students who were enrolled in schools before transition, in order to estimate the actual treatment effect on learning outcomes. Ansari (2020) also studies PPP programs in Punjab, Pakistan, researching whether they are effectively targeting districts with low enrolment, and which house- hold and parental characteristics predict whether a child attends a public school, a PPP school or a private school in Punjab. Rather than evaluating a single pro- gram, Ansari considers all of Punjab’s PPP programs, which includes voucher, sub- sidy and contract schools. He finds that PPP schools appear to have been located in districts where high shares of children are out of school—with the Foundation As- sisted Schools (FAS)—a subsidy program—and the New Schools Programme (NSP) being located in districts with the highest shares of out-of-school children. Prior to the period of our review, two studies (Malik 2010; Barrera-Osorio and Raju 2015) also find that FAS has successfully increased access to education for poorer children, suggesting that the FAS program design is somewhat effective in terms of reaching poor children—by removing fees at private schools near where poor children live— and handing them greater choice. Ansari (2020) finds no evidence that overall, children in PPP schools come from relatively more or less advantaged households than public school students. Children attending NSP schools and PSSP schools appear to be the most disadvantaged. While children attending EVS schools are more likely to be female, there is no gender differ- ential in any of the other PPP programs. He discusses the design objective of each PPP, noting that the FAS and NSP programs are most successful in targeting poorer dis- tricts, while the EVS program requires multiple private schools to be available in a lo- cality, hindering its ability to reach rural, underserved districts. All the PPP programs evaluated cost the same or less than the government cost-per-child of $9 per month, potentially offering substantial cost savings to the state (see supplementary online ap- pendix table A2). None of the studies look in detail at the cost differences between gov- ernment and PPP schools, but the most obvious is teacher salaries, with civil service salaries being substantially higher than private school teacher salaries. Notably, all these PPPs were established under the Punjab Education Foundation (PEF), a quasi- independent statutory body, rather than the Ministry of Education itself (PSSP, how- ever, was subsequently moved out of PEF and into the Ministry of Education). 114 The World Bank Research Observer, vol. 39, no. 1 (2024) Chile Neilson et al. (2019) study the voucher program in Chile, leveraging an im- portant change in the design of the program. They highlight the importance of endogenous response by schools to policy in determining outcomes. The 2008 Subsidio Escolar Preferencial (SEP) law raised the school voucher amount and elim- inated fees for the poorest 40 percent of students. This led to improved learning for Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 the poorest students, in particular where schools faced greatest competition. Neil- son et al. show that schools cut quality when they do not face competition and that PPP (voucher) policy design can influence school quality through its effects on competitive incentives. Specifically, they found that eliminating top-up fees and targeted vouchers for poorer communities contributed to improved academic out- comes. The introduction of a higher-value voucher for poorer students raised com- petition in these neighborhoods by giving poorer families the choice of schools that were previously unaffordable to them. A review of voucher programs by Baum (2018) further suggests that school vouchers can be an effective way of expanding education, particularly in underserved areas. He argues that vouchers make most sense where there are specific constraints to enrolment, such as household finance barriers, combined with an existing supply of private schools that have space to expand. Case Studies on Access: Sindh, Pakistan and Uganda In the context of recent findings from Punjab, Pakistan, Liberia and Chile, two ear- lier studies are important to discuss. Evaluations of the Uganda Secondary Education (USE) program and the Private Schooling in Rural Sindh (PPRS) program shed some light on how PPPs could be used by governments to increase access to education at a lower unit cost than the public sector. First, the Ugandan USE program. Established in 2007, the Ministry of Education and Sports’ USE program offered state subsidies for pupils enrolled at eligible private schools. Schools were eligible if they were based in sub-counties in which there were no public secondary schools, where those government schools were overcrowded, or where pupils must travel very long distances to reach the closest government school. Any student scoring more than 28 in the primary leaving exam was eligible for the subsidy. By 2016, nearly a third of all secondary school students in Uganda were en- rolled in PPP schools. The program had a positive effect on enrolment, and students in PPP schools performed better in English and mathematics than students in non- PPP private schools (Barrera-Osorio et al. 2020). The cost of the subsidy—47,000 Ugandan shillings per term—is much less than the cost of educating a child in a pub- lic school and did not require upfront infrastructure investment by the government. Crawfurd et al. 115 However, data are not available showing the full cost of the program to the government. Secondly, the PPRS program. Unlike Uganda, PPRS is not managed directly by the government. Similar to the Punjab PPPs described earlier, it is run by the Sindh Ed- ucation Foundation (SEF), a quasi-governmental agency which provides pre-vetted private entities with a per-student subsidy to operate no-fee, non-selective primary Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 schools along with additional educational resources. A rigorous study found that the intervention had a strong effect on enrolment and learning outcomes (Barrera- Osorio et al. 2022). While the subsidy paid to private schools is low—less than half the cost of government schooling—high start-up and administrative costs meant that after the first year the total cost to government was broadly equivalent to the cost of educating a child in the public system. Economies of scale meant that non-subsidy costs fell from 70 percent of total costs to less than 30 percent of total costs over three years. Increasing access to education was a core goal of both programs. Having an- nounced free, universal secondary education, the government in Uganda deployed three strategies to increase the number of places available to students: double-shifting in government schools; a program of construction of Government-aided “seed” sec- ondary schools in sub-counties without secondary schools; and the subsidy for pri- vate schools. One of the stated objectives of PPRS was to increase access to schooling in marginalized areas. In both cases, the subsidies paid to private schools to educate children are lower than the government cost per child. After high set-up costs, it seems likely that PPRS will deliver education at a lower cost than the government. Whether the total cost to the Ugandan government of the USE PPP is lower is more difficult to disentangle, as administrative costs are not readily available, although all indica- tions suggest that it is. This cost-effectiveness is further increased by infrastructure savings: since both programs primarily aimed to increase access in underserved ar- eas, double shifting and expansion of existing schools had limited potential. The cost to the government of building new schools is not included in this cost-effectiveness analysis and is likely to be substantial. Finally, the different management arrangements of these two PPPs is interest- ing to note. PPRS was outsourced to the Sindh Education Foundation (SEF), a semi- autonomous organization with a sizable management structure. The USE PPP was managed directly by staff at the Ministry of Education and Sports. Analysis on the benefits and drawbacks of outsourcing PPPs would be a valuable contribution to this debate. The PPPs in Uganda and Sindh suggest that, where access to education remains a challenge, PPPs can be a cost-effective way to rapidly increase enrolment. However, despite some seemingly positive attributes, challenges remained. While PPRS contin- ues, the USE PPP was terminated by the Government of Uganda, with reports that it did not have broad stakeholder buy-in (Aslam et al. 2017). 116 The World Bank Research Observer, vol. 39, no. 1 (2024) Discussion and Conclusions In this paper we have reviewed the evidence on the effectiveness of private schools, private school chains, and public-private partnerships. There is now considerable ev- idence that the average effectiveness of private schools is moderately higher than that of public schools. But this does not lead to obvious policy conclusions. Our analysis suggests that chains of private schools are unlikely to ever gain significant scale in Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 most school systems, especially when compared with the already-scaled non-chain private school sector. One limitation of this paper is our focus on learning outcomes in the definition of school effectiveness. The revealed preference of millions of parents for paid schools demonstrates that they gain enough value from private schools to justify the cost— and this may be partly due to other factors, such as shorter travel distances (Carneiro et al. 2022). The evidence on public-private partnerships is inconclusive and more limited in volume, but shows that overall the private school advantage has not translated into PPPs that reliably improve quality. But, PPPs can be a way to rapidly increase access to education at lower cost than through government provision, as shown by evidence from Pakistan (Barrera-Osorio et al. 2017) and Uganda (Barrera-Osorio et al. 2016; Crawfurd 2017). However, this cost-saving is typically driven almost entirely by lower teacher salaries and may not be a palatable political option: for example, in India, where a subsidy PPP operates at massive scale, private school teachers are mobiliz- ing to be paid salaries at parity with their counterparts in public schools. Therefore, expecting teachers in private schools subsidized by the government to work at much lower wages can create significant political tension and may not be feasible in the long term. A similar story has played out in Kenya with lower-paid contract teach- ers mobilizing for equal pay. Nonetheless, as countries grapple with SDG targets for universal secondary and universal pre-primary education, particularly in the wake of the COVID-19 economic crisis, these PPPs might be considered. The two papers from Liberia highlight the risks entailed in contracting private op- erators to run public schools, risks which may not be outweighed by relatively mod- est learning gains and risks which urge great caution. The government capacity re- quired to commission, contract and monitor private operators is substantial and not easily within the capabilities even of much better resourced countries. The experi- ence in Punjab, Pakistan, taken as a whole, paints a slightly more positive picture of the potential of PPPs. More research is needed to unpack whether the PSSP has actually increased enrollment or improved learning, but there is some evidence to suggest that the longer-standing NSP and FAS programs have successfully targeted poor areas and enrolled poor children. All the PPPs in Punjab are cost-effective, in that they cost the government the same or—in some cases substantially—less than the cost of educating a child in the public sector. PPPs seem to lower per-unit costs of Crawfurd et al. 117 education through two mechanisms: lower teacher salaries (Kim et al. 1999; Andrabi et al. 2008; Muralidharan and Kremer 2008; Barrera-Osorio et al. 2017; Barton et al. 2017), which may not be a viable long-term strategy for political reasons; and a lower cost of establishing a PPP school compared to a government school (Kim et al. 1999). The 2008 reform in Chile, studied by Neilson et al. (2019) offers some further insights into how a voucher program might be designed to mitigate risks of segregation and Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 to improve outcomes—namely through eliminating top-up fees and targeting poorer children with higher vouchers. For a PPP to succeed, the public side of the PPP needs to be strong. Good private sector performance in a particular country is no indication that the Ministry of Ed- ucation in that country can run the procurement, monitoring and evaluation, and overall governance of a public-private partnership more effectively than it manages its own schools. The government needs the capability to commission and regulate its private partners, and to hold them accountable for their performance. There are a range of conditions that must be in place to ensure these theoretical relationships ac- tually work as expected. For instance, the actual PPP contracts themselves need to be well designed, and the government needs to have the capacity to ensure they can be implemented and enforced. In some instances this has been achieved. Malik (2010) highlights strong monitoring and evaluation through a clear framework as being the main mechanism through which the Punjab Education Foundation improved student outcomes. In contrast, the actions of some providers in LEAP in Liberia had negative consequences which may have generated negative spillovers for the broader educa- tion system. These consequences were only picked up in the RCT and, in the absence of an RCT, may have continued. Expensive RCTs cannot compensate for weak mon- itoring systems. The negative consequences of providers’ actions in PSL underscore the need for robust monitoring by the government (Romero et al. 2020). Finally, the politics of PPPs are challenging. As discussed above, cost-efficiencies driven by lower teacher salaries may not be sustainable in the long-term. And public perceptions of PPPs can be polarizing. On the one hand, constituencies who believe that PPPs are a threat to public education may resist them. On the other hand, where the public do not perceive a PPP as a government intervention, lack of recognition for the government’s role in the provision of public education can reduce political sup- port. In Uganda, a factor in the decision to phase out the private secondary school PPP was that the important contribution the program plays in helping to deliver sec- ondary education in Uganda was not fully understood by many key stakeholders, from government to parents (O’Donoghue et al. 2018). The role of private schools, chains, and PPPs in global education remains deeply controversial. The Global Partnership for Education (GPE) prohibits funding for “commercial” providers of education,6 unless circumstances are exceptional, and the IFC has announced a freeze7 of all investments in for-profit K-12 schools. At the same time, surveys of senior officials in developing country Ministries of Education show that they are least content with the advice that they receive from donors on 118 The World Bank Research Observer, vol. 39, no. 1 (2024) private schools, relative to other less politicized areas such as curricula, examinations, teachers, and textbooks (Crawfurd et al. 2021b). A promising avenue for public en- gagement with private schools may be trying to address market failures such as those around credit and information, rather than through direct subsidy to specific schools (Andrabi et al. 2016). Important research questions remain. One argument around the reliance on pri- Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 vate school provision is that they may undermine public support for public financing of education. This question is explored by Emmerich Davies in the wake of a voucher experiment in Andhra Pradesh, India. There is some limited evidence that attending private schools does not substantially change political engagement (Andrabi et al. 2010; Davies 2022; Crawfurd et al. 2021a), but further evidence on this question would be welcome. Funding Information This work was supported by the UNESCO Global Monitoring Report and the Bill and Melinda Gates Foundation. Notes Center for Global Development, London, UK. Lee Crawfurd is corresponding author and can be reached at lcrawfurd@cgdev.org. 1. UNESCO Institute for Statistics via World Bank Open Data. Accessed October 24, 2022. https://data.worldbank.org/indicator/SE.PRM.PRIV.ZS 2. In appendix A, we include meta-analysis results including studies with other kinds of quasi- experimental designs. 3. https://www.cdcgroup.com/en/our-impact/key-data/Accessed October 5, 2020. 4. Disclosure: One of us, Susannah, was part of the founding board of the Global Schools Forum. 5. In addition to these 9,000+ fee-paying schools, BRAC operates more than 40,000 no-fee com- munity schools which are de facto a non-government parallel school system. 6. https://www.devex.com/news/education-advocates-claim-groundbreaking-victory-on-latest- gpe-decision-95110 7. https://financialservices.house.gov/news/documentsingle.aspx?DocumentID=406486 References Abdo, M., A. Assomull, J. Kapoor, and H. Wong. 2019. Private Schools for Public Goods Exploring the po- tential of privately-run schools to benefit societies. L.E.K. Consulting and Jacobs Foundation. Abdulkadiro˘glu, A., J. D. Angrist, P. D. Hull, and P. A. Pathak. 2016. “Charters without Lotter- ies: Testing Takeovers in New Orleans and Boston.” American Economic Review 106: 1878–920. https://doi.org/10/f8vsjd Alcott, B., and P. Rose. 2016. “Does Private Schooling Narrow Wealth Inequalities in Learn- ing Outcomes? Evidence from East Africa.” Oxford Review of Education 42: 495–510. https://doi.org/10.1080/03054985.2016.1215611 Crawfurd et al. 119 Alvi, S., A. N. Butt, and A. Khurshid. 2013. “Beaconhouse School System (BSS) Restructuring.” Asian Journal of Management Cases 10: 113–24. https://doi.org/10.1177/0972820113493687 Andrabi, T., N. Bau, J. Das, and A. I. Khwaja. 2010. Are Bad Public Schools Public “Bads”? Test Scores and Civic Values in Public and Private Schools. Bulletin of Education & Research.. Andrabi, T., J. Das, and A. I. Khwaja. 2008. “A Dime a Day: The Possibilities and Limits of Private School- ing in Pakistan.” Comparative Education Review 52 (3): 329–55. Andrabi, T., J. Das, and A. I. Khwaja. 2016. “Delivering Education: A Pragmatic Framework for Improv- Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 ing Education in Low-Income Countries,” In Handbook Of International Education. Andrews, J. 2016. School performance in multi-academy trusts and local authorities. Education Policy Institute. Andrews, J., and N. Perera. 2017. The Impact of Academies on Educational Outcomes. Education Policy Institute. Angrist, J. D., S. M. Dynarski, T. J. Kane, P. A. Pathak, and C. R. Walters. 2010. “Inputs and Impacts in Charter Schools: KIPP Lynn.” American Economic Review 100: 239–43. https://doi.org/10/fkpcsg Angrist, J. D., P. A. Pathak, and C. R. Walters. 2013. “Explaining Charter School Effectiveness.” American Economic Journal: Applied Economics 5 (4): 1–27. https://doi.org/10/gftrqb. Ansari, A. H., 2020. “Cream Skimming? Evaluating the Access to Punjab’s Public-private Partnership Programs in Education.” International Journal of Educational Development 72. Ashley, L. D, C. Mcloughlin, M. Aslam, J. Engel, J. Wales, S. Rawal, R. Batley, G. Kingdon, S. Nicolai, and P. Rose., et al. 2014. “The Role and Impact of Private Schools in Developing Countries (Education Rigorous Literature Review).” Department for International Development. Aslam, M., S. Rawal, and S. Saeed. 2017. Public Private Partnerships (PPPs) in Education in Developing Countries: A Rigorous Review of the Evidence. Education Partnerships Group, London. Assomull, A., M. Abdo, and R. Pelley. 2015. Driving Grades, Driving Growth: How Private Capital in Edu- cation is Increasing Access, Inspiring Innovation and Improving Outcomes. Parthenon-EY. Azam, M., G. Kingdon, and K. B. Wu. 2016. “Impact of Private Secondary Schooling on Cognitive Skills: Evidence from India.” Education Economics 24: 465–80. https://doi.org/10/gftrp3 Bagde, S., D. Epple, and L. Taylor. 2022. “The Emergence of Private High Schools in India: The Impact of Public-private Competition on Public School Students.” Journal of Public Economics 215: 104749. https://doi.org/10.1016/j.jpubeco.2022.104749 Barrera-Osorio, F. 2007. “The Impact of Private Provision of Public Education: Empirical Evidence from Bogota’s Concession Schools.” World Bank Policy Research Working Paper. Barrera-Osorio, F., D. S. Blakeslee, M. Hoover, L. Linden, D. Raju, and S. P. Ryan. 2017. “Delivering Educa- tion to the Underserved through a Public-private Partnership Program in Pakistan.” National Bureau of Economic Research. Barrera-Osorio, F., P. d. Galbert, J. Habyarimana, and S. Sabarwal. 2016. Impact of Public-Private Part- nerships on Private School Performance: Evidence from a Randomized Controlled Trial in Uganda. World Bank. Barrera-Osorio, F., and D. Raju. 2015. “Evaluating the Impact of Public Student Subsidies on Low-Cost Private Schools in Pakistan.” Journal of Development Studies 51: 808–25. https://doi.org/10/gftrw8 Barrera-Osorio, F., P. D. Galbert, J. Habyarimana, and S. Sabarwal. 2020. “The Impact of Public- Private Partnerships on Private School Performance: Evidence from a Randomized Controlled Trial in Uganda.” Economic Development and Cultural Change 68: 429–69. Barrera-Osorio, F., D. S. Blakeslee, M. Hoover, L. Linden, D. Raju, and S. P. Ryan. 2022. “Delivering Ed- ucation to the Underserved through a Public-Private Partnership Program in Pakistan.” The Review of Economics and Statistics 104: 399–416. https://doi.org/10.1162/rest_a_01002 120 The World Bank Research Observer, vol. 39, no. 1 (2024) Barry, K. B. 2019. “Report of the UN Special Rapporteur on the Right to Education: The Implementation of the Right to Education and Sustainable Development Goal 4 in the Context of the Growth of Private Actors in Education.” Barton, N., T. Bold, and J. Sandefur. 2017. “Measuring Rents from Public Employment: Regression Dis- continuity Evidence from Kenya,” CGD Working Paper. Baum, D. R. 2018. Private school vouchers in developing countries: A survey of the evidence (No. 6j7qp), SocArXiv. Center for Open Science. Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Baum, D. R., H. Abdul-Hamid, and H. T. Wesley. 2018. “Inequality of Educational Opportunity: The Relationship between Access, Affordability, and Quality of Private Schools in Lagos, Nigeria.” Oxford Review of Education 44: 459–75. https://doi.org/10.1080/03054985.2017.1421153 Bhatta, P., and T. Pherali. 2017. Nepal: Patterns of Privatisation in Education A case study of low-fee private schools and private chain schools. Education International. Bidwell, K., and L. Watine. 2014. “Exploring Early Education Programs in Peri-urban Settings in Africa: Final Report.” Innovations for Poverty Action. Brandt, K.. 2019. “Private Beats Public: A Flexible Value-Added Model with Tanzanian School Switch- ers.” Economic Development and Cultural Change, in 2022. Carneiro, P., J. Das, and H. Reis. 2022. “The Value of Private Schools: Evidence from Pakistan.” RISE Working Paper Series. 22/091. https://doi.org/10.35489/BSG-RISE-WP_2022/091 Cheng, A., C. Hitt, B. Kisida, and J. N. Mills. 2017. ““No Excuses” Charter Schools: A Meta-Analysis of the Experimental Evidence on Student Achievement.” Journal of School Choice 11: 209–38. https://doi.org/10.1080/15582159.2017.1286210 Crawfurd, L. 2018. “Contracting out Schools at Scale: Evidence from Pakistan.” RISE Working Paper 022. Crawfurd, L. 2017. “School Management and Public–Private Partnerships in Uganda.” Journal of African Economies 26: 539–60. https://doi.org/10/gcmcn5 Crawfurd, L., and A. Alam. 2022. “Contracting Out Schools at Scale: Evidence from Pakistan.” Education Economics 1–17. https://doi.org/10.1080/09645292.2022.2113859 Crawfurd, L., and A. Lipcan. 2018. “Learning in Lagos: Comparing Student Achievement in Bridge, Pub- lic, and Private Schools.” Crawfurd, L., D. Patel, and J Sandefur. 2021a. “Heterogeneous Effects of Low-Cost Private Schools: Ex- perimental Evidence from Delhi.” Unpublished Working Paper, Center for Global Development. Crawfurd, L., S. Hares, A. Minardi, and J. Sandefur. 2021b. “Understanding Education Policy Prefer- ences: Survey Experiments with Policymakers in 35 Developing Countries.” CGD Working Paper Cen- ter for Global Development.. Damera, V. K. 2017. “Choice for the Poor or Poor Choice? Experimental Evidence from Implementation of India’s School Choice Policy.” DPhil Thesis Blavatnik School of Government, University of Oxford. Davies, E. 2022. “The Lessons Private Schools Teach: Using a Field Experiment To Understand the Ef- fects of Private Services on Political Behavior.” Comparative Political Studies 001041402211151. https://doi.org/10.1177/00104140221115178 Dixon, P., A. J. Egalite, S. Humble, and P. J. Wolf . 2019. “Experimental Results from a Four-year Targeted Education Voucher Program in the Slums of Delhi.” World Development 124: 104644. https://doi.org/10.1016/j.worlddev.2019.104644 Eble, A., C. Frost, A. Camara, B. Bouy, M. Bah, M. Sivaraman, J. Hsieh, C. Jayanty, T. Brady, P. Gawron, S. Vansteelandt, P. Boone, and D Elbourne. 2020. “How Much Can We Remedy Very Low Learn- ing Levels in Rural Parts of Low-income Countries? Impact and Generalizability of a Multi-pronged Para-teacher Intervention from a Cluster-randomized Trial in the Gambia.” Journal of Development Economics 102539. https://doi.org/10.1016/j.jdeveco.2020.102539 Crawfurd et al. 121 Economic Policy and Research Centre. 2016. “Evaluation of the PEAS Network under the Uganda Sec- ondary Education (USE) Programme.” Eigbiremolen, G. O. 2019. “Estimating Private School Premium for Primary School Children in Ethiopia: Evidence from Individual-level Panel Data.” Progress in Development Studies 1464993419889708. https://doi.org/10.1177/1464993419889708 Eigbiremolen, G. O., J. E. Ogbuabor, and C. S. Nwambe. 2020. “Estimating Private School Effects for School Children in Peru: Evidence from Individual-level Panel Data.” Journal of International Devel- Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 opment 32: 131–49. https://doi.org/10.1002/jid.3438 Elacqua, G., D. Contreras, F. Salazar, and H. Santos. 2011. “The Effectiveness of Private School Fran- chises in Chile’s National Voucher Program.” School Effectiveness and School Improvement 22: 237–63. https://doi.org/10/cn53fw Evans, D. K., and F. Yuan. 2022. “How Big Are Effect Sizes in International Education Studies?” Educa- tional Evaluation and Policy Analysis 44: 532–40. https://doi.org/10.3102/01623737221079646 Fazzio, I., A. Eble, R. L. Lumsdaine, P. Boone, B. Bouy, P.-T. J. Hsieh, C. Jayanty, S. Johnson, and A. F Silva. 2021. “Large Learning Gains in Pockets of Extreme Poverty: Experimental Evidence from Guinea Bissau.” Journal of Public Economics 199: 104385. https://doi.org/10.1016/j.jpubeco.2021.104385 Featherston, S., and D. Ferreira. 2017. “The Business of Education in Africa.” Technical Report, Caerus Capital. Friedman, M., and R. D. Friedman. 1962. Capitalism and Freedom. University of Chicago Press. Gabaix, X. 2009. “Power Laws in Economics and Finance.” Annual Review of Economics 1: 255–94. https://doi.org/10.1146/annurev.economics.050708.142940 Gupta, S. 2020. “Do Preschools Add ‘Value’? Evidence on achievement gaps from rural India (Working Paper Series).” Department of Economics, University of Sussex Business School. Harma, J. 2016. “School Choice in Rural Nigeria? The Limits of Low-fee Private Schooling in Kwara State.” Comparative Education https://doi.org/10/gftssg. Hutchings, M., and B. Francis. 2017. “Chain Effects: The Impact of Academy Chains on Low-income Students.” The Sutton Trust. Istance, D., and A. Paniagua. 2019. Learning to Leapfrog: Innovative Pedagogies to Transform Education. Brookings Institution. Johnson, D., and P. J. Hsieh. 2019. “A Longitudinal Study of Learning, Progression, and Personal Growth in Sierra Leone.” Technical Report, Rising Academies Network. Joshi, P. 2020. “Do Private Schools Improve Public School Quality or Increase Stratifica- tion?” International Journal of Educational Development 77: 102219. https://doi.org/10.1016/ j.ijedudev.2020.102219 Kenny, C., J. Kalow, B. Leo, and V. Ramachandran. 2018. “Comparing Five Bilateral Development Fi- nance Institutions and the IFC.” Center for Global Development, Policy Paper. Kim, J., H. Alderman, and P. F. Orazem. 1999. “Can Private School Subsidies Increase Enroll- ment for the Poor? The Quetta Urban Fellowship Program.” World Bank Economic Review. https://doi.org/10/gftrz5. Kingdon, G. G. 2020. “The Private Schooling Phenomenon in India: A Review.” Journal of Development Studies 0: 1–23. https://doi.org/10.1080/00220388.2020.1715943 Karopady, D. D., 2014. “Does School Choice Help Rural Children from Disadvantaged Sections? Evidence from Longitudinal Research in Andhra Pradesh.” Economic and Political Weekly 49 (51): 46–53. LaRocque, N. 2008. Public-private partnerships in basic education: An international review. CfBT Education Trust. Leaver, C., R. Lemos, and D. Scur. 2019. “Measuring and Explaining Management in Schools: New Ap- proaches Using Public Data.” Policy Research Working Paper, World Bank. 122 The World Bank Research Observer, vol. 39, no. 1 (2024) Lipcan, A., L. Crawfurd, and B. Law. 2018. “Learning in Lagos: Comparing Student Achievement in Bridge, Public, and Private Schools.” Oxford Policy Management. Malik, A. B. 2010. Public–Private Partnerships in Education Lessons Learned from the Punjab Education Foun- dation. Asian Development Bank. Monarrez, T. E. 2022. “School Attendance Boundaries and the Segregation of Public Schools in the U.S.” American Economic Journal: Applied Economics. https://doi.org/10.1257/app.20200498 Mundy, K., and F. Menashy. 2013. “Investing in Private Education for Poverty Alleviation: The Case Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 of the World Bank’s International Finance Corporation.” International Journal of Educational Develop- ment. https://doi.org/10/f5vxd4. Muralidharan, K., and M. Kremer. 2008. “Public-Private Schools in Rural India.” In School Choice In- ternational: Exploring Public-Private Partnerships, edited by R. Chakrabarti and P.E. Peterson. The MIT Press. https://doi.org/10.7551/mitpress/9780262033763.003.0005 Muralidharan, K., and V. Sundararaman. 2015. “The Aggregate Effect of School Choice: Evi- dence from a Two-Stage Experiment in India.” Quarterly Journal of Economics 130: 1011–66. https://doi.org/10/f7nvd2 Neilson, C., C. Allende, and F. Gallego. 2019. “Approximating the Equilibrium Effects of Informed School Choice.” Princeton University Industrial Relations Section, Working Paper. Nguyen, Q. T., and D. Raju. 2015. “Private School Participation in Pakistan.” Lahore Journal of Economics 46. O’Donoghue, J., L. Crawfurd, J. Makaaru, P. Otieno, and R. Perakis. 2018. A review of Uganda’s Universal Secondary Education Public Private Partnership programme. Education Partnerships Group (EPG). Patel, D., and J. Sandefur. 2020. “A Rosetta Stone for Human Capital (CGD Working Paper).” Center for Global Development, Working Paper. Patrinos, H. A., F. Barrera-Osorio, and J. Guaqueta. 2009. The Role and Impact of Public-Private Partner- ships in Education. The World Bank. DOI: 10.1596/978-0-8213-7866-3. Pritchett, L. 2014. “The Risks to Education Systems from Design Mismatch and Global Isomorphism: Concepts, with Examples from India,” WIDER Working Paper 2014/039. Romero, M., and J. Sandefur. 2019. Beyond Short-term Learning Gains: (No. CGD Working Paper 521). Romero, M., J. Sandefur, and W. A. Sandholtz. 2020. “Outsourcing Education: Experimental Evidence from Liberia.” American Economic Review 110: 364–400. https://doi.org/10.1257/aer.20181478 Siddiqui, N. 2017. “Socio-economic Segregation of Disadvantaged Children between Schools in Pakistan: Comparing the state and Private Sector.” Educational Studies 43: 391–409. https://doi.org/10.1080/03055698.2016.1277139. Singh, A. 2015a. “How Standard Is a Standard Deviation? A Cautionary Note on Using SDs To Com- pare across Impact Evaluations in Education [WWW Document].” World Bank Development Impact. Accessed March 6, 2015. http://blogs.worldbank.org/impactevaluations/how-standard-standard- deviation-cautionary-note-using-sds-compare-across-impact-evaluations. Singh, A. 2015b. “Private School Effects in Urban and Rural India: Panel Estimates at Primary and Sec- ondary School Ages.” Journal of Development Economics 113: 16–32. https://doi.org/10/gftrsm Smith, W. C., and T. Baker. 2017. “From Free to Fee: Are for-Profit, Fee-Charging Private Schools the Solution for the World’s Poor?” RESULTS Educational Fund. Srivastava, P., and C. Noronha. 2016. “The Myth of Free and Barrier-free Access: India’s Right to Education Act – private Schooling Costs and Household Experiences.” Oxford Review of Education. https://doi.org/10/gftssw. Woodworth, J. L., M. E. Raymond, C. Han, Y. Negassi, and W. P. Richardson. 2017. Charter Management Organizations. Center for Research on Educational Outcomes. Crawfurd et al. 123 Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Supplementary Online Appendix The Impact of Private Schools, School Chains and PPPs in Developing Countries Lee Crawfurd , Susannah Hares, and Rory Todd Appendix A: Additional Meta-Analysis Results There results include studies using quasi-experimental designs other than value- added models. This includes three using fixed-effects, one using propensity-score matching, and one using a weak instrumental variable design. Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Figure A1.1. Meta-Analysis of Private School Effects on Learning (Including All Quasi- Experimental Studies) Note: Where we have extracted multiple results from a study, the effect given is the fixed effect for that study in a model with fixed effects for each study. Figure A1.2. Differences in Management Quality between Public and Private Schools Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: Data is from the PISA survey, following Leaver et al. (2019). Figure A1.3. Development Finance Institution (DFI) Investments in Education Note: The figure on the left shows total education investments by six major Development Finance Institutions between 2012 and 2016. The figure on the right shows the distribution of investments by the largest DFI in education—the World Bank IFC. The source for the figure on the left is Kenny et al. (2018). The source for the figure on the right is our analysis of IFC submissions to IATI. Figure A1.4. Few Chains are Larger than the Average Government School District Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: The number of government schools per district was obtained for respective government websites. Data on schools for most private chains was obtained from the Global Schools Forum (GSF). We also added data on some non-GSF members—Samata and Chaudhary Group from Bhatta and Pherali (2017) and Beaconhouse from Alvi et al. (2013) and http://www.educators.edu.pk/ (accessed October 5, 2020). Figure A1.5. In Most Systems, the Median Number of Schools in Chains is Very Small and the Percentage of Schools that are Part of a Chain is Also Small Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: Data for UK preschools is from daynurseries.co.uk. For US charter schools from CREDO. The mean chain size is reported rather than the median for US charters. Data for UK academies is from gov.uk. For India is from Gray Matters Capital. For Chile from datos.mineduc.cl. For Nairobi from IPA (Mukuru) and CapPlus. For Kampala from CapPlus. For Lagos and Accra, from IPA. Figure A1.6. Per-Pupil Cost in the Liberia PPP. Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Note: Data comes from budgets submitted to Social Finance, who managed the pool of funds that paid providers the per pupil subsidy. One operator (Stella Maris) did not provide budget data. Numbers do not include the cost of teaching staff borne by the Ministry of Education. The red line represents USD 50 per pupil. Source: Romero and Sandefur 2019. Table A1.1. Studies on Different PPP Types Aslam et al. (2017)[2009–2016] This review[2017–2020] Contract 3 3 Subsidy 9 Voucher 8 1 Other 2 1 Total 22 5 Note: This table shows the number of studies on different types of public-private partnerships included in the earlier review (Aslam et al. 2017) and in this review. Crawfurd et al. 1 Table A1.2. Public-Private Partnerships in Punjab, Pakistan Program Year launched Main features No. of schools Cost per child ($) Education Voucher 2006 Tuition vouchers for 1,650 3.30–6.60 Scheme children in poor areas to attend eligible private schools Downloaded from https://academic.oup.com/wbro/article/39/1/97/7115930 by World Bank and IMF user on 02 February 2024 Foundation Assisted 2005 Per student subsidies 3,700 3.30–9 Schools provided directly to private schools that meet criteria New Schools Program 2008 Establish new schools 2,404 3.30–9 where there is no public school within 1km PSSP 2016 Outsourcing public schools 4,276 4.20 to private operators (organizations)–3.30 (individuals) Note: This table summarizes key features of the four main public-private partnership schemes in Punjab, Pak- istan. Source: Punjab Education Foundation website (accessed 28/8/20) and Crawfurd (2018). Exchange rates as of August 28, 2020.